使用 Matplotlib 使两个标记共享相同的标签并同时在图例中保留图块的黑边
Make two markers share the same labels and simultaneously keep the black edge of patch in legend using Matplotlib
我应该如何修改我的以下代码，使两个标记共享相同的标签，同时在图例中保留补丁的黑边？这是代码（我已经绘制了两次条形图以分离边缘颜色和填充颜色）和绘制的图片。我想做第二张图的标记。
import matplotlib
matplotlib.rcParams['pdf.fonttype'] = 42
matplotlib.rcParams['ps.fonttype'] = 42
import matplotlib.pyplot as plt
x_lable = ['3', '4', '5', '6', '7']
width = 0.3
Data1 = [96, 99, 100, 100, 100]
Data2 = [94, 96, 95, 96, 95]
Data3 = [48, 43, 42, 42, 37]
data_3 = [2228, 2621, 3165, 3761, 3763]
data_4 = [3895, 5670, 7354, 8999, 10731]
data_5 = [4355, 6373, 8279, 10105, 12018]
xcoordinate = [1, 3, 5, 7, 9]
xcoordinate_1 = [xcoordinate[0] + 0 * width, xcoordinate[1] + 0 * width, xcoordinate[2] + 0 * width, xcoordinate[3] + 0 * width, xcoordinate[4] + 0 * width]
xcoordinate_2 = [xcoordinate[0] + 1 * width, xcoordinate[1] + 1 * width, xcoordinate[2] + 1 * width, xcoordinate[3] + 1 * width, xcoordinate[4] + 1 * width]
xcoordinate_3 = [xcoordinate[0] + 2 * width, xcoordinate[1] + 2 * width, xcoordinate[2] + 2 * width, xcoordinate[3] + 2 * width, xcoordinate[4] + 2 * width]
xcoordinate_4 = [xcoordinate[0] + 3 * width, xcoordinate[1] + 3 * width, xcoordinate[2] + 3 * width, xcoordinate[3] + 3 * width, xcoordinate[4] + 3 * width]
xcoordinate_5 = [xcoordinate[0] + 4 * width, xcoordinate[1] + 4 * width, xcoordinate[2] + 4 * width, xcoordinate[3] + 4 * width, xcoordinate[4] + 4 * width]
fig = plt.figure()
# plot ax1
ax1 = fig.add_subplot(111)
ax1.bar(xcoordinate_3, data_3, width=width, label='GGG', color='none', edgecolor='orange', hatch='\\\')
ax1.bar(xcoordinate_4, data_4, width=width, label='SSS', color='none', edgecolor='blue', hatch='---')
ax1.bar(xcoordinate_5, data_5, width=width, label='KKK', color='none', edgecolor='orangered', hatch='///')
ax1.set_ylabel('BBB')
ax1.set_xlabel('AAA')
ax1.bar(xcoordinate_3, data_3, width=width, label='GGG', color='none', edgecolor='black')
ax1.bar(xcoordinate_4, data_4, width=width, label='SSS', color='none', edgecolor='black')
ax1.bar(xcoordinate_5, data_5, width=width, label='KKK', color='none', edgecolor='black')
handles1, labels1 = ax1.get_legend_handles_labels()
order = [0, 1, 2]
plt.legend([handles1[idx] + handles1[idx+3] for idx in order], [labels1[idx] for idx in order], loc=(2.5/10, 1.08), frameon=True, ncol=3, shadow=False, framealpha=1, labelspacing=1.05).get_frame().set_edgecolor('black')
# plot ax2
ax2 = ax1.twinx()
ax2.plot([i + 3 * width for i in xcoordinate], Data3, label='GGG', color='orange', linestyle='-', marker='p')
ax2.plot([i + 3 * width for i in xcoordinate], Data2, label='SSS', color='orangered', linestyle='-', marker='s')
ax2.plot([i + 3 * width for i in xcoordinate], Data1, label='KKK', color='limegreen', linestyle='-', marker='^')
ax2.set_ylabel('CCC')
plt.xticks([i + 3 * width for i in xcoordinate], x_lable)
handles2, labels2 = ax2.get_legend_handles_labels()
order = [0, 1, 2]
plt.legend([handles2[idx] for idx in order], [labels2[idx] for idx in order], loc=(2.5/10, 1.005), frameon=True, ncol=3, shadow=False, framealpha=1, labelspacing=1.05).get_frame().set_edgecolor('black')
plt.show()
绘制的图片：
我想要的：
当两个制造商共享同一个标签时，我想保留补丁的黑边。
官方指南中的第三个例子就是你想要的。要组合每个处理程序，请将它们设置为元组形式。如果不想让每个handler都堆叠起来，可以加个配置。
from matplotlib.legend_handler import HandlerLineCollection, HandlerTuple
#plt.legend([handles1[idx] + handles1[+3] for idx in order], [labels1[idx] for idx in order], loc=(2.5/10, 1.08), frameon=True, ncol=3, shadow=False, framealpha=1, labelspacing=1.05).get_frame().set_edgecolor('black')
plt.legend([(handles1[idx],handles2[idx]) for idx in order], [labels2[idx] for idx in order], loc=(2.5/10, 1.005), frameon=True, ncol=3, shadow=False, framealpha=1, labelspacing=1.05,handler_map={tuple: HandlerTuple(ndivide=None)}).get_frame().set_edgecolor('black')
看来，我们不能简单地使用HandlerTuple class任意分组句柄（或者我不够聪明）。然而，当ndivide定义的条目更多时，key又从位置1开始，所以我们可以这样写：
....
from matplotlib.legend_handler import HandlerTuple
...
order = [0, 1, 2]
plt.legend([(handles1[idx], handles2[idx], handles1[idx+3]) for idx in order], [labels1[idx] for idx in order], handler_map={tuple: HandlerTuple(ndivide=2)}, loc=(0.1, 1.04), frameon=True, ncol=3, shadow=False, framealpha=1, labelspacing=1.01, handlelength=5).get_frame().set_edgecolor('black')
....
为每个图例条目将 3 个键分组到 2 个位置。这给出了以下输出：
我还增加了手柄长度...惊喜...handlelength=5 因为否则，我们将看不到太多填充图案。
我应该如何修改我的以下代码，使两个标记共享相同的标签，同时在图例中保留补丁的黑边？这是代码（我已经绘制了两次条形图以分离边缘颜色和填充颜色）和绘制的图片。我想做第二张图的标记。
import matplotlib
matplotlib.rcParams['pdf.fonttype'] = 42
matplotlib.rcParams['ps.fonttype'] = 42
import matplotlib.pyplot as plt
x_lable = ['3', '4', '5', '6', '7']
width = 0.3
Data1 = [96, 99, 100, 100, 100]
Data2 = [94, 96, 95, 96, 95]
Data3 = [48, 43, 42, 42, 37]
data_3 = [2228, 2621, 3165, 3761, 3763]
data_4 = [3895, 5670, 7354, 8999, 10731]
data_5 = [4355, 6373, 8279, 10105, 12018]
xcoordinate = [1, 3, 5, 7, 9]
xcoordinate_1 = [xcoordinate[0] + 0 * width, xcoordinate[1] + 0 * width, xcoordinate[2] + 0 * width, xcoordinate[3] + 0 * width, xcoordinate[4] + 0 * width]
xcoordinate_2 = [xcoordinate[0] + 1 * width, xcoordinate[1] + 1 * width, xcoordinate[2] + 1 * width, xcoordinate[3] + 1 * width, xcoordinate[4] + 1 * width]
xcoordinate_3 = [xcoordinate[0] + 2 * width, xcoordinate[1] + 2 * width, xcoordinate[2] + 2 * width, xcoordinate[3] + 2 * width, xcoordinate[4] + 2 * width]
xcoordinate_4 = [xcoordinate[0] + 3 * width, xcoordinate[1] + 3 * width, xcoordinate[2] + 3 * width, xcoordinate[3] + 3 * width, xcoordinate[4] + 3 * width]
xcoordinate_5 = [xcoordinate[0] + 4 * width, xcoordinate[1] + 4 * width, xcoordinate[2] + 4 * width, xcoordinate[3] + 4 * width, xcoordinate[4] + 4 * width]
fig = plt.figure()
# plot ax1
ax1 = fig.add_subplot(111)
ax1.bar(xcoordinate_3, data_3, width=width, label='GGG', color='none', edgecolor='orange', hatch='\\\')
ax1.bar(xcoordinate_4, data_4, width=width, label='SSS', color='none', edgecolor='blue', hatch='---')
ax1.bar(xcoordinate_5, data_5, width=width, label='KKK', color='none', edgecolor='orangered', hatch='///')
ax1.set_ylabel('BBB')
ax1.set_xlabel('AAA')
ax1.bar(xcoordinate_3, data_3, width=width, label='GGG', color='none', edgecolor='black')
ax1.bar(xcoordinate_4, data_4, width=width, label='SSS', color='none', edgecolor='black')
ax1.bar(xcoordinate_5, data_5, width=width, label='KKK', color='none', edgecolor='black')
handles1, labels1 = ax1.get_legend_handles_labels()
order = [0, 1, 2]
plt.legend([handles1[idx] + handles1[idx+3] for idx in order], [labels1[idx] for idx in order], loc=(2.5/10, 1.08), frameon=True, ncol=3, shadow=False, framealpha=1, labelspacing=1.05).get_frame().set_edgecolor('black')
# plot ax2
ax2 = ax1.twinx()
ax2.plot([i + 3 * width for i in xcoordinate], Data3, label='GGG', color='orange', linestyle='-', marker='p')
ax2.plot([i + 3 * width for i in xcoordinate], Data2, label='SSS', color='orangered', linestyle='-', marker='s')
ax2.plot([i + 3 * width for i in xcoordinate], Data1, label='KKK', color='limegreen', linestyle='-', marker='^')
ax2.set_ylabel('CCC')
plt.xticks([i + 3 * width for i in xcoordinate], x_lable)
handles2, labels2 = ax2.get_legend_handles_labels()
order = [0, 1, 2]
plt.legend([handles2[idx] for idx in order], [labels2[idx] for idx in order], loc=(2.5/10, 1.005), frameon=True, ncol=3, shadow=False, framealpha=1, labelspacing=1.05).get_frame().set_edgecolor('black')
plt.show()
绘制的图片：
我想要的：
当两个制造商共享同一个标签时，我想保留补丁的黑边。
官方指南中的第三个例子就是你想要的。要组合每个处理程序，请将它们设置为元组形式。如果不想让每个handler都堆叠起来，可以加个配置。
from matplotlib.legend_handler import HandlerLineCollection, HandlerTuple
#plt.legend([handles1[idx] + handles1[+3] for idx in order], [labels1[idx] for idx in order], loc=(2.5/10, 1.08), frameon=True, ncol=3, shadow=False, framealpha=1, labelspacing=1.05).get_frame().set_edgecolor('black')
plt.legend([(handles1[idx],handles2[idx]) for idx in order], [labels2[idx] for idx in order], loc=(2.5/10, 1.005), frameon=True, ncol=3, shadow=False, framealpha=1, labelspacing=1.05,handler_map={tuple: HandlerTuple(ndivide=None)}).get_frame().set_edgecolor('black')
看来，我们不能简单地使用HandlerTuple class任意分组句柄（或者我不够聪明）。然而，当ndivide定义的条目更多时，key又从位置1开始，所以我们可以这样写：
....
from matplotlib.legend_handler import HandlerTuple
...
order = [0, 1, 2]
plt.legend([(handles1[idx], handles2[idx], handles1[idx+3]) for idx in order], [labels1[idx] for idx in order], handler_map={tuple: HandlerTuple(ndivide=2)}, loc=(0.1, 1.04), frameon=True, ncol=3, shadow=False, framealpha=1, labelspacing=1.01, handlelength=5).get_frame().set_edgecolor('black')
....
为每个图例条目将 3 个键分组到 2 个位置。这给出了以下输出：
我还增加了手柄长度...惊喜...handlelength=5 因为否则，我们将看不到太多填充图案。